brycewang-stanford Reproducibility & Supplementary Material (ectheory-replication-and-data-policy)
When to trigger
- Long proofs/derivations exceed the 50-page Article ceiling and need a home
- You want your Monte Carlo results to be reproducible and your code clean
- Preparing the final-manuscript files and unsure what ET requires vs forbids in the Supplement
- Wondering whether ET has a formal data/code deposit requirement (it does not — see below)
ET's reality: no mandatory archive, a Supplementary-Material policy
Unlike empirical economics journals, ET is theory-focused and states no mandatory formal data/code/replication archive on its author-instructions pages (recorded as ABSENT as of 2026-06-01; 待核实). Note: the JAE Data Archive belongs to a different journal (Journal of Applied Econometrics) and does not apply to ET. What ET does operate is a voluntary online Supplementary Material policy:
- Only critical, already-reviewed material may be posted online with the article — long mathematical derivations, supporting lemmas, and extra tables/figures.
- It is submitted with the final manuscript as a separate, clearly labeled file.
- It is not copyedited or typeset.
- It may not introduce new, unseen material — everything in it must have gone through review.
This is the standard ET route for the long technical proofs and simulation overflow that do not fit in the main paper under the 50-page ceiling.
Building the Supplement well
- Make it a self-contained document (compiles independently) that shares the main paper's assumption numbering and notation, so cross-references resolve.
- Put here: complete proofs of auxiliary lemmas, lengthy derivations, additional Monte Carlo DGPs and tables, and robustness simulations — all already part of what referees saw.
- Do not smuggle new theorems, new claims, or unreviewed results into the Supplement.
Reproducible computation (good practice, not a deposit mandate)
- Keep one runnable script per table/figure (or a master script) regenerating all simulation output.
- Fix and report random seeds; record replication counts, sample sizes, and the full DGP.
- Pin the computational environment (package/version notes) so results regenerate.
- A clean, reproducible numerical appendix builds referee trust even absent a formal archive.
Checklist
- Confirmed ET has no mandatory data/code archive; not budgeting for a deposit step (待核实)
- Overflow proofs/derivations placed in a separate, clearly labeled Supplement file
- Supplement contains only already-reviewed material; no new results introduced
- Supplement is self-contained and shares main-paper notation/numbering
- Simulation code runnable; seeds, reps, and DGP reported
- Main Article respects the 50-page ceiling; overflow lives in the Supplement
Anti-patterns
- Assuming an AEA/JAE-style mandatory archive applies to ET (it does not)
- Introducing a new theorem or claim only in the non-reviewed Supplement
- A Supplement that cannot be read without the main paper's lost context
- Non-reproducible simulations (unreported seeds / replication counts)
- Padding the main Article past 50 pages instead of using the Supplement
What belongs where: a theory-venue routing table
With no mandatory data/code archive, the reproducibility decision is mostly where each piece of technical content lives.
| Content | Main Article | Online Supplement | Forbidden |
|---|---|---|---|
| Main theorem + proof sketch | yes | — | — |
| Long proof of an auxiliary lemma | sketch only | full proof | — |
| Extra Monte Carlo DGPs / robustness tables | headline result | overflow tables | — |
| A brand-new theorem not yet refereed | — | — | yes (no new results) |
The hard rule: the Supplement carries only already-reviewed material and ships as a separate labeled file with the final manuscript. Confirm conventions against the author guidelines (mandatory archive recorded ABSENT as of 2026-06-01; 待核实).
Worked vignette and the reproducibility fixes
A paper proves a uniform convergence rate. Its maximal-inequality proof (Lemma 2) and full size/coverage grid exceed the page ceiling and move to a self-contained Supplement reusing the paper's numbering. Ship one master script that regenerates every cell with the environment pinned:
# reproducibility manifest accompanying the simulation Supplement
seed_master = 314159 # reported in table notes
reps = 20000 # per design cell
n_grid = [100,200,400,800]
p_grid = [5,10,25,50] # dimension growing with n
software = R 4.4.x; packages: <list with versions>
mapping = each table -> the script block that produces it
This builds referee trust, not a deposit mandate — there is no AEA/JAE-style archive obligation at ET. The fixes: "the auxiliary proof is missing" → include it in the reviewed Supplement during the review round, not for the first time post-review; "simulations not reproducible" → add the manifest above and a table-to-script mapping; "Supplement unreadable on its own" → make it compile independently against the numbered theorems.
Output format
【Archive requirement】none mandatory at ET (待核实)
【Supplement】separate labeled file, already-reviewed only? [Y/N]
【Self-contained】shares notation/numbering? [Y/N]
【No new material】confirmed? [Y/N]
【Reproducibility】seeds + reps + DGP + env recorded? [Y/N]
【Next step】ectheory-review-process